1997] STREAM ORGANIC MA I 1 hR BUDGETS 95

Temperature fluctuation and nitrate were highlycorrelated with their 2nd principle component.One difference between their analysis and ourswas that precipitation was highly correlated withtheir 1st principle component but didn't comeinto our analysis until the 3rd principle compo-nent. The lack of a temperature effect and thegreater importance of precipitation in the studyby Cushing et al. were probably results of themore limited latitudinal extent of their sites.

Conclusions

The 35 streams used in this study were notchosen to represent the range of physical con-ditions of streams throughout the world. Rather,they were chosen because of the availability ofdata on organic processes. However, wherecomparisons were possible, relationships amongphysical variables of the 35 streams were similarto the relationships among the same variablesin more extensive studies. Stream size, watertemperature, and precipitation appear to be themost important variables physically character-izing the streams. These variables set the tem-plate for organic processes occurring in streams.However, as becomes evident in analyses of or-ganic processes, the influence of these physicalprocesses is largely indirect through their effectson terrestrial vegetation.

Acknowledgements

This work was supported by the CoweetaLong Term Ecological Research grant from theNational Science Foundation. Bert Cushing andJudy Meyer provided helpful comments on anearlier version of this chapter.

Literature Cited

CUSHING, C. E., C. D. MCINTIRE, J. R. SEDELL, K. W.CUMMINS, G. W. MINSHALL, R. C. PETERSEN, ANDR. L. VANNOTE. 1980. Comparative study ofphysical-chemical variables of streams using mul-tivariate analyses. Archiv far Hydrobiologie 89:343-352.

DUNNE, T., AND L. B. LEOPOLD. 1978. Water in envi-ronmental planning. W. H. Freeman, San Francis-co.

LEOPOLD, L. B. 1994. A view of the river. HarvardUniversity Press, Cambridge, Massachusetts.

LEOPOLD, L. B., M. G. WOLMAN, AND J. R. MILLER.1964. Fluvial processes in geomorphology. W. H.Freeman, San Francisco.

LIKENS, G. E., AND F. H. BORMANN. 1995. Biogeo-chemistry of a forested ecosystem. Springer-Ver-lag, New York.

MORISAWA, M. 1968. Streams, their dynamics andmorphology. McGraw-Hill, New York.

PARK, C. C. 1977. World-wide variations in hydraulicgeometry exponents of stream channels: an anal-ysis and some observations. Journal of Hydrology33:133-146.

SCHLESINGER, W. H. 1991. Biogeochemistry: an anal-ysis of global change. Academic Press, San Diego,California.

STRAHLER, A. N. 1957. Quantitative analysis of wa-tershed geomorphology. Transactions of theAmerican Geophysical Union. 38:913-920.

STRAHLER, A. N. 1975. Physical geography. John Wi-ley and Sons, New York.

SWIFT, L. W, G. B. CUNNINGHAM, AND J. E. DOUGLASS.

1988. Climatology and hydrology. Pages 35-55 inW. T. Swank and D. A. Crossley (editors). Foresthydrology and ecology at Coweeta. Springer-Ver-lag, New York.

WEBSTER, J. R., AND J. L. MEYER. 1997. Stream organicmatter budgets—introduction. Pages 5-13 in J. R.Webster and J. L. Meyer (editors). Stream organicmatter budgets. Journal of the North AmericanBenthological Society 16:3-161.

WOLMAN, M. G., AND R. GERSON. 1978. Relativescales of time and effectiveness of climate in wa-tershed geomorphology. Earth Surface Processes3:189-208.

A comparison of primaryproduction in stream ecosystems

GARY A. LAMBERTI

Department of Biological Sciences,University of Notre Dame,

Notre Dame, Indiana 46556 USA

ALAN D. STEINMAN

Department of Ecosystem Restoration,South Florida Water Management District,

West Palm Beach, Florida 33416 USA

The objective. of this paper is to identify phys-ical, chemical, and biological variables thatmight help explain the wide range of primaryproduction observed in streams from a varietyof biomes and locations throughout the world.We used regression approaches to search forpredictive, statistical relationships that might re-veal how aquatic, riparian, and watershed vari-

96 J. R. WEBSTER AND J. L. MEYER (EDITORS) [Volume 16

ables are associated with differences in primaryproduction among 30 streams from the originaldata set (see Webster and Meyer 1997) for whichprimary production was measured.

Primary production may be defined generallyas the conversion of solar energy to reducedchemical energy, or specifically as the amount oforganic matter formed from inorganic carbon byphotosynthetic organisms during a specifiedtime interval (Bott 1996). The total amount ofcarbon fixed during the interval is the gross pri-mary production (GPP). Common intervalsused to describe GPP are a day or a year. Be-cause plants also respire to drive their cellularmetabolism, some of this fixed energy is lost asCO2; this process is termed autotrophic respi-ration (R k). The remaining fixed carbon allocat-ed to biomass represents the net primary pro-duction (NPP). Therefore,

GPP = NPP + RA

In streams, primary producers generally areassociated with benthic substrates, and includealgae, cyanobacteria, bryophytes, and vascularmacrophytes, although algae are usually themajor producers. Large rivers may develop aself-sustaining phytoplankton assemblage,termed potamoplankton, that contributes to pri-mary production (Hynes 1970). Benthic algaealso are called periphyton, but periphvton ac-tually is a mixture of algae, bacteria, protists,fine detritus, and other materials that attach tosubmersed surfaces. In practice, it is almost im-possible to separate the metabolic rates of au-totrophs from heterotrophs in studies of loticprimary production. Thus, these metabolic mea-surements also include the respiration of organ-isms other than plants, termed heterotrophicrespiration (R,i ). For this reason, many investi-gators prefer to call the respiration term com-munity or ecosystem respiration (R 1) to ac-knowledge the heterotrophic component (cf.Gregory 1983).

The focus of this paper is on annual GPP fromour sample of 30 streams. We made the follow-ing predictions concerning the patterns of GPP,some of which were based on tenets of the rivercontinuum concept (Vannote et al. 1980).

1) GPP is low in low-ord-r (headwater) streamsbecause of light lim:ration by riparian cano-py, and thus GPP is negatively related to per-cent canopy cover;

GPP increases as stream size increases butdeclines in large, deep rivers;this stream size-GPP relationship also is re-flected in correlates of stream size (e.g., dis-charge, gradient, channel width, watershedarea);GPP declines with increasing latitudebecause of lower annual light inputs and av-erage water temperatures;GPP is negatively related to annual precipi-tation because of increased riparian vegeta-tion (and greater shading) and more scour byfloods; and

6) GPP is positively related to nutrient concen-trations.

Methods

Annual GPP estimates were available for 30streams, mostl y from the northern hemisphereand predominately at latitudes between 30 and50° (see Webster and Meyer 1997 for a completelist). To facilitate comparison across different bi-omes, we classified the streams into 5 broad cat-egories based on climate and forest type: desert

and arid land streams, deciduous foreststreams, boreal coniferous forest streams, mon-tane coniferous forest streams, and other

streams (Fig. 1). The "other" group includedblackwater streams, grassland streams, and tun-dra streams (see Webster and Me yer 1997).

Measurements of GPP, especially those madein chambers, generally were for periphvton at-tached to rocks (sometimes termed epilithon).However, some sites included harvest of mac-rophytes, measurement of water column pro-ductivity, or inclusion of bryophytes in incuba-tion chambers, where appropriate. The frequen-cy of GPP estimates varied with site, rangingfrom 2 (e.g., ice-free period at extreme northernlatitudes) to 12 (monthly at several sites) mea-surements. Measurements of plant standingcrop (e.g., chlorophyll a or biomass) were not asreadily available as GPP estimates, and thus wedid not include standing crop in our analysis.

Several different, but widely accepted, meth-ods were used to estimate benthic GPP of thestreams. More than half of the studies (n = 17)measured 0, evolution from sediments placedin closed chambers. Seven studies monitored isitu diel 02 curves, and one study (Rattiest-, _IkeSprings, Washington) used the diel CO, curve.Oxygen measurements were converted to car-

6

>-.5c`.1

Y'4

0

3

0

2E-2-

,(;),0

1997] STREAM ORGANIC MATTER BUDGETS 97

<C1ND< <u3000V)Y

¢« Z - 0 02(5MQ-1-2ZZclaccOc00coccico 0- < v)

COCOCOYCOCO000<0001-0000

2 2 2 co

CECCESCCITCC0 0 0 0 0 0CC 0 'cr 0

M 0

<<<YZ0(5 Y <<ZCrQ- co

Desert Deciduous Boreal Coniferous OtherFIG. 1. Annual gross primary production in streams used in this analysis. Acronyms are defined by Webster

and Meyer (1997).

bon units based on atomic weight (C = 0, x0.375). Incubations with radiolabeled carbon( 14C) were conducted in 5 studies. For "C incu-bations and the diel CO, curve, GPP was as-sumed to be 2X the NPP ( 14C and diel CO, ap-proaches are believed to estimate NPP; Schin-dler 1978). The photosynthetic quotient used byinvestigators ranged from 1.0 to 1.35.

We analyzed the data using linear regressionapproaches. GPP was the dependent (response)variable in all regressions. Several independent(predictor) variables were chosen from the en-tire data set for examination. We categorizedthese variables into 3 groups: 1) geographic/cli-matic variables = latitude, mean annual watertemperature, and precipitation; 2) basin/chan-nel variables = stream order, watershed area,discharge, gradient, and channel width; 3) re-source variables = nutrient concentrations andlight. We had sufficient data to examine the fol-lowing nutrients: SRP (soluble reactive phos-phorus), TP (total phosphorus), NH 4-N (am-monium-nitrogen), and NO .,-N (nitrate-nitro-gen). Unfortunately, there were insufficient dataon irradiance for direct inclusion in the analysis;instead, we used percent riparian canopy cover(generally reported) as an inverse surrogate forlight (see below). Variables whose values rangedover several orders of magnitude (e.g., GPP, dis-

charge, watershed area) were logrn-transformedprior to analysis. In cases where only ranges ofvalues were available (e.g., nutrient concentra-tions), the mean of maximum and minimumvalues was used in the regression. Simple linearregression (SLR) was first performed using eachpredictor variable and the response variable(log 10 GPP). In this analysis, we realize thatmany of the variables were correlated (e.g., area,order, and discharge) but we were searching forstatistical patterns in the data set. Finally, a step-wise multiple linear regression was conductedusing all significant (p < 0.05) predictors fromSLRs and the response variable (log 10 GPP).

Results

GPP in the 30 streams varied by over 4 ordersof magnitude, ranging from 3.5 to 5400 g C m-2y-' (Fig. 1). Streams in arid regions of the USsuch as the Great Basin (Rattlesnake Springs,Washington; Deep Creek, Idaho) and the desertSouthwest (Sycamore Creek, Arizona) (groupedas "desert") had the highest GPP. GPP was low-est for streams in deciduous forests of the east-ern USA (Bear Brook, New Hampshire; SatelliteBranch and Hugh White Creek, North Caroli-na). Average GPP for all streams was 560 g Cm -2 y- I.

B

• •••

•

•

••

r2.0.142• p-0.044

120 30 40 50 60 70 80 90

Latitude0 2 4 6 8 10 12 14 16 18 20

Temperature (°C)

10000

1000 -

100 -

10 -

A

10i i 1 1

0 2 4 6 8Stream order

0 50 100 150 200 250 300Precipitation (cmly)

•i.■

♦

••

D

DeciduousBorealDesertConiferousOther

?-0.074p=0.153

•♦

■

••

98

J. R. WEBSTER AND J. L. MEYER (EDITORS)

[Volume 16

F IG. 2. Regressions of GPP (g C m- 2 y-') on 3 geographic/climatic variables (A-C) and 1 basin/channelvariable (D).

Geographic/climatic variables

GPP showed no relationship with latitude(Fig. 2A), even though the range of latitudesrepresented was fairly broad. GPP increasedsignificantly with mean annual temperature (r2= 0.142, p = 0.044; Fig. 2B) and declined withtotal precipitation (r2 = 0.206, p = 0.013; Fig.2C). However, the 3 North American desertstreams (low precipitation, high temperature)strongly influenced both regressions; if those 3points were removed from the analysis, therewas no significant relationship for any climaticvariable.

Channel variables

Increases in annual GPP were strongly asso-ciated with increasing watershed area (Fig. 3A).Area explained over 40% of the variation in GPP(r2 = 0.416, p < 0.001). GPP also was positivelyassociated with discharge (r2 = 0.168, p = 0.027)(Fig. 3B) ?Ad negatively associated with gradi-ent (r2 = 0.208, p = 0.015) (Fig. 3C). Discharge

and gradient are, in part, a function of water-shed area (increasing discharge and decliningaverage gradient with increasing area). Therewas no significant relationship between GPPand stream order (Fig. 2D) or channel width(Fig. 3D). In fact, the greatest variation in GPPwas expressed in the smallest streams (but thesewere also the most frequently measured).

Resource variables

Of the nutrients examined, only soluble re-active phosphorus (SRP) was significantly relat-ed to GPP (r2 = 0.382, p < 0.001). This relation-ship was steep and asymptotic, with the desertstreams having the highest SRP and GPP (Fig.4A). A 2nd-order function also was applied tothe SRP-GPP regression, given the obvious non-linear relationship between the variables, andthe r2 improved to 0.479 (GPP = 1.700 + 0.023SRP - 6.026 SRP'). Small streams draining de-ciduous forest had the lowest S'::P and GPP. Nosignificant relationships were observed between

• • ■••

• •

501 I 1 1 1

0 1 0 20 30 40

Channel width (m)

10000

1000 -a_a.

100 -

10 -

■ • • fOno

See

0.001 0.01 0.1Gradient (m/m)

•

C ••r2=0.208p-0.015

%•

0.0001

••

•B t7-0.168p.0.027

A r2-0.416•P<°'°°1

••• ■•■ •

•

••••• •100 -

•00

1 1 1 I 1 I 1

10 1 102to' 104 105 106 leWatershed area (ha)

10 1102 105 104 105 10 6 107Discharge (Us)

10000

1000 -

a_LsridCC

10 -

••

♦• •■

•••

D • ••

V•

••

/2-0.005p-0.748

•

DeciduousBorealDesertConiferousOther

1997] STREAM ORGANIC MATTER BUDGETS 99

FIG. 3. Regressions of GPP (g C m- 2 y- 1 ) on 4 basin/channel variables.

GPP and total phosphorus, NI-1 4-N, or NO,-N(Fig. 4B—D). Indeed, the 2 forms of nitrogeneach explained less than 2% of the variation inGPP.

The lack of irradiance data prompted us touse percent riparian canopy cover as an inversesurrogate for light. Values used in the regressionwere estimates of maximum canopy closure.GPP was significantly related to percent canopycover (r2 = 0.291, p = 0.003; Fig. 5). Desertstreams tended to have the lowest canopy coverand highest GPP whereas deciduous foreststreams had the highest canopy and lowest GPP.

log ioGPP = 0.717 + 0.689 log 10 Area— 0.494 log 10 Discharge + 0.3871og,„ SRI'

= 17.2, p < 0.001, n = 27)

These 3 predictors combined explained 70% ofthe variation in GPP, of which watershed areaexplained 42%, discharge explained an addi-tional 21%, and phosphorus explained the final7%. A multiple regression with all 7 variablesincluded in the model explained only a smalladditional amount of variation in GPP (r2 =0.73).

DiscussionMultiple regression analysis

The 7 significant predictors from the simplelinear regressions (precipitation, watershedarea, discharge, percent canopy, gradient, tem-perature, and SRP) were used in a stepwisemultiple regression. Three significant predictorswere included in the final step of the regression:

Autochthonous production is a complex func-tion of the geomorphic, fluvial, and riparian set-ting of the stream. As recognized 20 y ago byMinshall (1978) and verified by the data set thatwe have analyzed in this paper, lotic GPP varieswidely across biomes, climatic regions, andstream ecosystems. In our analysis, small to me-

J. R. WEBSTER AND J. L. MEYER (EDITORS)

10000

[Volume 16

• •B • r2.0.006

1000 •0.

r2.0.382p<0.001 p.0.797

100 t ••

•■•

DeciduousBorealDesert

■ ••ti

••

•• Coniferous

10 - ♦ Other

C•

•

•r2.0.018p0 603

10000

•i •

••

•

•

1000 -

100 -

10 -

100

0 50 100 150 200 250 300Soluble reactive phosphorus (gg/L)

I I 1

0 20 40 60 80 100 120Total phosphorus (ug/L)

0 20 40 60 80 100 120 0 1000 2000 3000Ammonium (nil.)

Nitrate (ug/L)

FIG. 4. Regression of GPP (g C m -2 y- l ) on 4 resource variables—nutrient concentrations.

dium-sized streams with a limited canopy (e.g.,desert streams) tended to have high benthicGPP, presumably because light did not limitphotosynthesis. In contrast, small streams inheavily forested areas had low GPP (e.g., moun-tain streams in Oregon and North Carolina).

10000

1000

••

• •r2=0.291p=0.003

•100 -•;•♦

• • •• •

•

• •• Deciduous

10 ■ Boreal• Desert •• Coniferous 41♦ Other

0 20 40 60 80 100

% Canopy

FIG. 5. Regression of GPP (g C m- 2 y-') on 1 re-source variable—% overhanging riparian canopy.

Larger, deep rivers (e.g., Ogeechee River, Geor-gia; Moisie River, Quebec) had moderate levelsof benthic GPP, possibly because of reducedshading by vegetation offset by some light at-tenuation by the dissolved and suspended loadof water. These patterns generally are consistentwith predictions of the river continuum concept(Vannote et al. 1980) as tested and modified byothers (Cushing et al. 1983, Minshall et al. 1983).Part of this consistency in results may stemfrom the fact that some of the same streamsused to test the river continuum concept werealso included in our analysis. However, the re-sults in the present paper are based on a broad-er sampling of streams than the river continu-um, suggesting that these patterns are robust.

We originally predicted that GPP would below in small streams because of light limitationby riparian canopy. This prediction was sup-ported by an analysis of discharge (Fig. 3B) butnot by the analysi' of stream order (Fig. 2D).GPP varied over 3 orders of magnitude in both1st- and 2nd-order streams. Thus, local riparian

1997] STREAM ORGANIC MA 11ER BUDGETS 101

conditions appeared to override the predictedrelationship of GPP to stream order (Vannote etal. 1980), a pattern supported by the negativerelationship between % canopy and GPP (Fig.5). Although canopy data are not as directlylinked to primary production as irradiance, theyappear to be a relatively good surrogate. Instudies that have examined GPP versus irradi-ance in streams of different orders, a strong anddirect relationship exists (Bott et al. 1985, Nai-man et al. 1987).

Our 2nd prediction was that GPP would in-crease with stream size, but would decline againin large rivers. The first part of this predictionwas supported to some extent, but there wasonly a modest decline in GPP in large rivers.Discharge was a significant, but weak, linearpredictor of GPP. If the desert streams are dis-regarded, GPP appeared to reach an asymptoteat a discharge of about 1 m 3 /s (Fig. 3B). Thus,larger rivers appear to be as productive as me-dium-sized streams. The absence of a signifi-cant relationship between GPP and stream or-der may be due to: 1) a sampling bias towardlow-order streams in the regression; 2) thebroad range of local conditions reflected in low-order streams whereas larger rivers may inte-grate regional conditions; and 3) the subjectivityinherent in assigning stream order (Gordon etal. 1992). For example, because of differences inhydrologic regime and drainage network form,the discharges of similar order streams can varywidely in different geographic areas (Richards1982).

Watershed area was the best single predictorof GPP, perhaps because watershed size is acoarse-grain measure that summarizes manyfeatures relevant to lotic primary production.For example, nutrient inputs from terrestrialsources should increase with watershed size;but because discharge also increases, nutrientconcentrations may either increase or decrease.Streams of arid regions (with large watershed:stream size ratios) have high levels of insolationand warm water that can stimulate GPP and R A.

For 2 geographic sites, GPP was available for arange of stream sizes (5 sites in the Matamek/Moisie river systems of Quebec; 5 sites in theMackenzie River system of Oregon). For the Ma-tamek/Moisie rivers, GPP increased withstream size and was the highest in the 9th-orderMoisie River. In Oregon, GPP increased withstream size except between the 5th-order Look-

out Creek and the 7th-order Mackenzie River,where GPP was about equal. The 6th-orderOgeechee River, a Georgia blackwater stream,had much higher GPP than smaller streams inforested mountains of North Carolina. A limi-tation of this data set, however, is the lack ofmeasurements from very large rivers, whereGPP may in fact decline.

Our 3rd prediction was that the stream size-GPP relationship would also be reflected in cor-relates of stream size (e.g., discharge, gradient,channel width). This prediction was generallysupported, although relationships were weakerthan for watershed area. GPP increased withdischarge and declined with gradient. However,there was no relationship between GPP andchannel width; width may be an insensitivemetric because it is dependent on channel formand site-specific conditions (Gordon et al. 1992).

We predicted that higher latitude and associ-ated declines in the length of the growing sea-son, in total solar input, and in mean annualwater temperature would reduce GPP. We foundno relationship between GPP and latitude, but aweak positive relationship between GPP and av-erage water temperature. Higher water temper-ature may stimulate autotrophic respiration(and thus GPP). Perhaps more importantly, tem-perature is positively related to insolation,which may explain why open-canopy desertstreams had the highest temperature and GPP.

Our 5th prediction was that GPP would benegatively associated with annual precipitationas a result of at least 2 possible mechanisms.First, higher precipitation should result in moreriparian vegetation and thus greater shading ofthe stream. Second, concentrated precipitationcould result in scouring of benthic algae byfloods. This prediction was supported in thatstreams with the highest precipitation in theirwatersheds tended to have the lowest GPP.However, higher discharge was positively relat-ed with GPP, and discharge was a significantpredictor in the multiple regression. The posi-tive influence of discharge on GPP may be re-lated to increased loading of nutrients into thestream with storm runoff (cf. McDiffett et al.1989, Stevenson 1990), although high dischargedoes not imply a variable hydrograph. Eachstream probably has a unique, complex relation-ship between hydrology, riparian vegetation,and benthic primary production.

Finally, we predicted that GPP would be pos-

102 J. R. WEBSTER AND J. L. MEYER (EDITORS) [Volume 16

itively associated with nutrient concentrations in tion (as SRP) and insolation (as canopy) alsothe water. This prediction assumes that primary were significant predictors of lotic GPP.production is limited more by nutrients than by Perhaps more than any other measurementother resources such as light (c.f. Hill and made in stream ecology, the method used has aKnight 1988, Steinman 1992, Rosemond 1993). critical influence on apparent rates of primarySRP was the only nutrient positively related to production (Bott et al. 1978, Marzolf et al. 1994).GPP in our analysis. This result is consistent For example, Marzolf et al. (1994) compared me-with the belief that phosphorus generally is tabolism measured from upstream—down-more limiting than nitrogen in freshwater eco- stream changes in DO with that measured fromsystems (Schindler 1978, Hecky and Kilham DO changes in chambers and found that 24-h1988); however, there are streams where algal community respiration rates were 3 to 4X lowergrowth clearly is limited by nitrogen (Grimm in chambers than in the whole-stream tech-and Fisher 1986, Lohman et al. 1991). In addi- nique. They attributed this difference to the fail-tion, the significance of dissolved inorganic nu- ure of chambers to include the metabolism oftrient concentration in streamwater is question- macrophytes, macrofauna, and organic sedi-able, as it is the nutrient concentration of tissue ments. Although we assume that geomorphol-that influences metabolism (cf. Dodds 1993). ogy, climate, and biota have stronger effects onFurthermore, high nutrient concentrations may metabolic rates than do differences in methods,be a result of low demand by autotrophs, there- this assumption has not been tested rigorouslyby resulting in an inverse relationship between and should be considered when interpretingGPP and nutrients. Thus, the relationship be- patterns of lotic GPP.tween lotic GPP and the absolute concentrationof any nutrient should be interpreted with cau- Conclusionstion.

We compared our results for streams with

In our analysis of lotic primary production,some past analyses done for lakes. Brylinsky we found high variation in GPP among differentand Mann (1973) analyzed lake productivity streams, especially among streams of the samedata from the International Biological Program apparent order. Concepts of stream function fre-and concluded that variables related to solar in- quently are based on stream order; our analysisput (most notably latitude) were the best pre- suggests that assumptions about GPP based ondictors of primary production for 55 lakes rang- order should be made with caution, at leasting from the tropics to the Arctic. However, they when generalizing over large scales or acrosssuggested that for a narrower range of latitudes, different hydrologic regimes. Within a specificnutrient loading could assume greater impor- river network, inferences based on stream ordertance. Later, when more nutrient data were appear to be fairly robust (e.g., Minshall et al.available, Schindler (1978) concluded that phos- 1983, Sheath et al. 1986, Naiman et al. 1987).phorus input was the best predictor of primary Streams draining large watersheds tended toproduction in the north temperate region (only have higher GPP than those draining smaller ar-a single tropical lake, and no southern hemi- eas. Many features, such as light, nutrient input,sphere lakes were analyzed) while acknowledg- temperature, hydrology, and others, may being the significance of solar input over a broader driving this general relationship. Our multiplelatitudinal range. Our analyses of streams sug- regression analysis suggests that to make agest that lotic GPP is a complex function of geo- rough prediction of annual GPP for a stream, atmorphology, hydrology, and resources for a minimum, measurements of watershed area,aquatic and riparian plants. Large-scale features mean annual discharge, and average SRP(e.g., latitude) had little influence in our analy- should be taken. Further refinement of this anal-sis, but unlike Brylinsky and Mann's (1973) ysis will be possible when additional informa-study, tropical streams were absent from our tion is available for many streams, including: 1)data set. Watershed size, which influences and

direct measurements of irradiance; 2) frequency

integrates many other features of river systems and intensity of flood disturbance; and 3) con-including local irradiance and discharge, sumption of autotrophs (either directly with as-emerged as the best predictor of GPP. Consis- says rr indirectly from herbivore density or bio-tent with the lake analyses, nutrient concentra- mass), which may have considerable influence

1997]

STREAM ORGANIC MATTER BUDGETS 103

on autotrophic standing crop and production(Lamberti and Moore 1984, Steinman 1996).

Although our analysis has revealed some gen-eral relationships that we hope are useful in oth-er lotic studies, every stream will tend to havea unique set of factors that controls the rate ofprimary production. Evaluation of site-specificobservational data combined with rigorous ex-perimentation to test specific hypotheses at ap-propriate scales will be the most powerfulmeans of revealing mechanisms that govern pri-mary production in a particular stream.

Acknowledgements

We thank Jack Webster and Judy Meyer fororganizing the workshop leading to this paper.Jack Webster patiently answered our manyquestions concerning the data set, and JudyMeyer made important suggestions on the anal-ysis. The comments of Rosemary Mackay andseveral reviewers improved the chapter. Thiswork was supported, in part, by grants from theNational Science Foundation (BSR-8907968) andthe Environmental Protection Agency (CR-820290-01 and CR-820290-02).

Literature Cited

Burr, T. L. 1996. Primary productivity and commu-nity respiration. Pages 533-556 in E R. Hauer andG. A. Lamberti (editors). Methods in stream ecol-ogy. Academic Press, San Diego, California.

BOTT, T. L., J. T. BROOK, C. E. CUSHING, S. V GREGORY,D. KING, AND R. C. PETERSEN. 1978. A compari-son of methods for measuring primary produc-tivity and community respiration in streams. Hy-drobiologia 60:3-12.

Borr, T. L., J. T. BROCK, C. S. DUNN, R. J. NAIMAN, R.W. OvINK, AND R. C. PETERSEN. 1985. Benthiccommunity metabolism in four temperate streamsystems: an interbiome comparison and evalua-tion of the river continuum concept. Hydrobiol-ogia 123:3-45.

BRYLINSKY, M., AND K. H. MANN. 1973. An analysisof factors governing productivity in lakes and res-ervoirs. Limnology and Oceanography 18:1-14.

CUSHING, C. E., K. W. CUMMINS, G. W. MINSHALL, ANDR. L. VANNOTE. 1983. Periphyton, chlorophyll-a,and diatoms of the Middle Fork of the SalmonRiver, Idaho. Holarctic Ecology 6:221-227.

DODDS. W. K. 1993. What controls levels of dissolvedrhosphate and ammonium in surface waters?Aquatic Sciences 55:132-142.

GORDON, N. D., T. A. MCMAHON, AND B. L. FINLAY-

SON. 1992. Stream hydrology: an introduction forecologists. John Wiley & Sons, Chichester, UK.

GREGORY, S. V. 1983. Plant-herbivore interactions instream systems. Pages 157-189 in J. R. Barnes and

W. Minshall (editors). Stream ecology. PlenumPress, New York.

GRIMM, N. B., AND S. G. FISHER. 1986. Nitrogen lim-itation in a Sonoran Desert stream. Journal of theNorth American Benthological Society 5:2-15.

HECKY, R. E., AND P. KILHAM. 1988. Nutrient limita-tion of phytoplankton in freshwater and marineenvironments: A review of recent evidence on theeffects of enrichment. Limnology and Oceanog-raphy 33:796-822.

HILL, W. R., AND A. W. KNIGHT. 1988. Nutrient andlight limitation of algae in two northern Califor-nia streams. Journal of Phycology 24:125-132.

HYNES, H. B. N. 1970. The ecology of running waters.University of Toronto Press, Toronto.

LAMBERTI, G. A., AND J. W MOORE. 1984. Aquatic in-sects as primary consumers. Pages 164-195 in V

Resh and D. M. Rosenberg (editors). The ecol-ogy of aquatic insects. Praeger Publishers, NewYork.

LOHMAN, K., J. R. JoNEs, AND C. BAYSINGER-DANIEL.1991. Experimental evidence for nitrogen limita-tion in an Ozark stream. Journal of the NorthAmerican Benthological Society 10:14-23.

MARZOLF, E. R., P. J. MULHOLLAND, AND A. D. STEIN-

MAN. 1994. Improvements to the diurnal up-stream-downstream dissolved oxygen changetechnique for determining whole-stream metab-olism in small streams. Canadian Journal of Fish-eries and Aquatic Sciences 51:1591-1599.

MCDIFFETT, W. E, A. W. BEIDLER, T. F. DOMINICK, AND

D. MCCREA. 1989. Nutrient concentration-stream discharge relationships during stormevents in a first-order stream. Hydrobiologia 179:97-102.

MINSHALL, G. W. 1978. Autotrophy in stream ecosys-tems. BioScience 28:767-771.

MINSHALL, G. W., R. C. PETERSEN, K. W. CUMMINS, T.BOTT, J. R. SEDELL, C. E. CUSHING, AND R. L.

VANNOTE. 1983. Interbiome comparison ofstream ecosystem dynamics. Ecological Mono-graphs 53:1-25.

NAIMAN, R. J., J. M. MELILLO, M. A. LOCK, T. E. FORD,

AND S. R. REICE. 1987. Longitudinal patterns ofecosystem processes and community structure ina subarctic river continuum. Ecology 68:1139-1156.

RICHARDS, K. 1982. Rivers, form and process in al-luvial channels. Methuen, London.

ROSEMOND, A. D. 1993. Interactions among irradi-ance, nutrients, and herbivores constrain a streamalgal community. Oecolrgia 94:585-594.

SCHINDLER, D. W. 1978. Factors regulating phyto-plankton production and standing crop in the

104 J. R. WEBSTER AND J. L. MEYER (EDITORS) [Volume 16

world's freshwaters. Limnology and Oceanogra-phy 23:478-486.

SHEATH, R. G., J. M. BURKHOLDER, J. H. HAMBROOK,A. M. HOGELAND, E. HOY, M. E. KANE, M. 0.MORISON, A. D. STEINMAN, AND K. L. VAN AL-sTyNE. 1986. Characteristics of softwater streamsin Rhode Island. III. Distribution of macrophyticvegetation in a small drainage basin. Hydrobiol-ogia 140:183-191.

STEINMAN, A. D. 1992. Does an increase in irradianceinfluence periphyton in a heavily-grazed wood-land stream? Oecologia 91:163-170.

STEINMAN, A. D. 1996. Effects of grazers on fresh-water benthic algae. Pages 341-373 in R. J. Ste-venson, M. L. Bothwell, and R. L. Lowe (editors).Algal ecology: freshwater benthic ecosystems,Academic Press, San Diego, California.

STEVENSON, R. J. 1990. Benthic algal community dy-namics in a stream during and after a spate. Jour-nal of the North American Benthological Society9:277-308.

VANNOTE, R. L., G. W. MINSHALL, K. W. CUMMINS, J.R. SEDELL, AND C. E. CUSHING. 1980. The rivercontinuum concept. Canadian Journal of Fisheriesand Aquatic Sciences 37:130-137.

WEBSTER, J. R., AND R. L. MEYER. 1997. Stream or-ganic matter budgets—introduction. Pages 5-13in J. R. Webster and J. L. Meyer (editors). Streamorganic matter budgets. Journal of the NorthAmerican Benthological Society 16:3-161.

Comparison of litterfallinput to streams

E. F. BENFIELDDepartment of Biology,

Virginia Polytechnic Institute and State University,Blacksburg, Virginia 24061 USA

Allochthonous organic matter is an importantsource of energy for many streams and the ma-jor energy source for woodland streams orstreams with well developed riparian corridorsof vegetation (e.g., Cummins et al. 1983). Litter-fall may be defined as allochthonous materialentering streams from riparian vegetation. Itmay include leaves and leaf fragments, floralparts, bark, wood (branches and twigs), conesand nuts, fruits, and other plant parts (Bray andGorham 1964). Litter may reach streams by di-rect fall or lateral movement (blowing or slidingdown the stream banks). The relative ary,ountsof material reaching streams by these 2 routes

vary considerably. Lateral movement may varywith wind patterns, aspect, bank slope, and oth-er site-specific factors (Wallace et al. 1992). Forexample, lateral movement accounted for about24% of total litter input to 4 southern Appala-chian streams (Webster et al. 1995), about 66%in a Douglas fir-hemlock forest stream in thewestern US (Sedell et al. 1982), but only about10% in a eucalyptus forest stream in Australia(Campbell et al. 1992). The composition of lit-terfall varies with vegetation type and location.As a general average, non-leaf litterfall for for-ests around the world is about 30% (Bray andGorham 1964) but may be up to 70% in someforests in southeastern Australia (Blackburn andPetr 1979, Briggs and Maher 1983).

In temperate deciduous forests, the bulk of lit-terfall occurs in autumn but material may con-tinue entering streams by lateral movement overthe remainder of the year. Needle-fall from co-niferous evergreen trees varies considerablywith species and location and ma y range fromdistinctly seasonal to irregular throughout theyear (Bray and Gorham 1964). Litterfall fromtropical wet forest trees and shrubs is usuallynon-synchronous and leaves enter streams rel-atively evenly over the entire year (Stout 1980).

In streams with broadly developed valleys orin lowland systems, litter may be entrainedfrom the floodplain as streams rise during pe-riods of increasing discharge (Cuffney 1988).Conversely, litter may be deposited on thefloodplain as streams retreat during falling hy-drographs (Post and de la Cruz 1977, Shure andGottschalk 1985). Floodplain entrainment/de-position cycles of litter during changing hydro-graphs may also occur in smaller, montanestreams (Wallace et al. 1992) and tundra streams(Peterson et al. 1986). Thus floodplairt areas maybe sources or sinks for litterfall depending onhydrodynamics, topography, sediment loads,and other factors (Cuffney 1988). In some flood-plain systems, litterfall may be largely pro-cessed on the floodplain and the resulting par-ticles entrained by streams during high flows(Smock 1990).

The objectives of this chapter are to summa-rize data on direct fall and lateral movement oflitter to streams that were included in the earliersite-description chapters, and to analyze wheth-er patterns of direct litterfall to these streamsmight be explained on the basis of local or spe-

STREAM ORGANIC MATTER BUDGETS

1

6

EDITED BYJ. R. WEBSTER AND JUDY L. MEYER

Made in United States of AmericaReprinted from J OURNAL OF THE NORTH AMERICAN BENTHOLOGICAL SOCIETY

Vol. 16, No. 1, March 1997© 1997 by The North American Benthological Society

J. N. Am. Benthol. Soc., 1997, 16(1)3-161r 1997 by The North American Benthological Society

Stream organic matter budgets' 2

EDITED BYJ. R. WEBSTER AND JUDY L. MEYER

Abstract. This analysis of organic matter dynamics in streams has 3 objectives: 1) to explore therelationships between physical characteristics of streams and their watersheds (climate, geomorphol-ogy) and stream organic matter dynamics using data from a broad geographic area; 2) to comparestream organic matter dynamics in a diverse arra y of streams in order to suggest determinants ofobserved patterns; and 3) to reveal deficiencies in currently available data on organic matter dynamicsin streams. Streams were included in this anal ysis not to represent the global diversity of streamtypes but because organic matter data were available. In the introductory chapter we describe thekinds of data included for each stream and provide brief descriptions of previously published organicmatter data for streams included in the comparative analysis but not described in individual chapters.The next 16 chapters present organic matter data for streams from North America, Europe, Australia,and Antarctica. Most of the streams represented are in the temperate zone of North America. Datapresented include climate and geomorphic variables and organic matter inputs, exports, and standingcrops. The chapters on individual streams are followed by 7 chapters analyzing physical features ofthese streams and specific components of the organic matter budgets. Stream size, water temperature,and precipitation were the most important variables setting the physical template for organic matterprocesses occurring in the streams. Watershed area was the best predictor of gross primary produc-tivity (GPP), which increased with increasing watershed area. Watershed area, discharge, and solublereactive phosphorus concentration explained 71% of the variation in GPP. Climate (latitude) andvegetation type were more important than stream order in predicting litter inputs across a broadgeographic range of streams, although, within a river basin, litterfall decreased with increasingstream order. Regression of benthic organic matter (BOM) and latitude and precipitation proveduseful in predicting BOM standing crop in streams at a continental scale, although BOM was alsorelated to channel characteristics such as gradient and woody debris. Benthic respiration increaseddramatically with increasing temperature (Q, o = 7.6), suggesting a response related not only tometabolism but also to changes in BOM quality in response to latitudinal shifts in vegetation. Ter-restrial and riparian vegetation was found to play an important role in regulating suspended partic-ulate organic matter (POM) concentration and export, with higher values observed in forested streamsand in lower gradient streams with extensive floodplains. Channel slope was the best predictor ofdissolved organic matter (DOM) concentration and export, probably because of its relationship withriparian wetlands and hydrologic flowpaths. In the final chapter, a synthesis of the organic matterbudgets, we reached two conclusions: 1) At a global level, stream organic matter dynamics are drivenprimarily by climate through its effect on terrestrial vegetation. 2) Despite significant progress inunderstanding organic matter processes in streams, many of the differences we found among streamsreflect omissions of important components of the budget, especially accurate measures of streambedarea, heterotrophic respiration, standing stock of fine BOM, and groundwater inputs of DOM.

Kett anrds: stream, organic matter, budget, primary production, litterfall, BOM, DOM, POM, res-piration.

Table of Contents

Introduction (J. R. Webster and J. L. Meyer)

5

Canada Stream: a glacial meltwater stream in Taylor Valley, South Victoria Land,Antarctica (D. M. McKnight and C. M. Tate)

14

' Individual chapters should be cited as follows: Author(s). 1997. Chapter title. Pages xxx—xxx in J. R. Websterand J. L. Meyer (editors). Stream organic matter budgets. Journal of the North American Benthological Society16:3-161.

= Reprint requests for individual chapters should be sent to the arpropriate senior author. Correspondenceconcerning the entire paper should be directed to: J. R. Webster, Department of Biology, Virginia PolytechnicInstitute and State University, Blacksburg, Virginia 24061 USA.

3